Task scheduling and resource management algorithm on reconfigurable platform

Abstract

The invention provides a task scheduling and resource management algorithm on a reconfigurable platform, and the algorithm comprises a reconfigurable FPGA platform, and comprises the following steps of starting to process a task; binding the task; binding the task to a processor; realizing the long sleep, when no idle black box exists, entering a long sleep state, and after long sleep is finished,entering a task binding state; realizing the short sleep, if no black box can finish the current task, entering a short sleep state, and entering the task binding state after the short sleep is finished; reconfiguring, when the number of times that the algorithm enters short sleep reaches a threshold value, entering a reconstruction state from the task binding state; executing the task, when oneblack box can execute the task, sending data to the black box, and starting an execution process of the task; and ending, after the task execution is ended, returning data. By means of the reconfigurable FPGA platform, the task execution process is accelerated, and the task execution time is effectively shortened.

Description

technical field [0001] The invention relates to the technical field of dynamic scheduling, in particular to a task scheduling and resource management algorithm on a reconfigurable platform. Background technique [0002] The scheduling algorithms on the reconfigurable platform can be divided into three categories according to their execution characteristics, as follows: [0003] (1) On-demand scheduling: On-demand scheduling is one of the simplest scheduling methods. In this method, the reconstruction is performed according to the request sent by the task actively. For the reconstruction request that cannot be satisfied immediately, the queue is used Cache reconstruction requests. When the application program calls a hardware task, the reconstruction request is sent to the operating system. If the hardware for executing the task is already on, the task will be executed directly; otherwise, if there are enough hardware resources on the platform to lay out the hardware task,...

AI Technical Summary

Problems solved by technology

In addition, information such as the hardware resources used by the task and the execution time on the hardware also need to be determined. In practical applications, usually only some embedded applications for special devices can meet these constraints at the same time. In addition, if there are multiple concurrent or parallel applications in the system, dynamic scheduling of an application alone cannot give reasonable scheduling results, so the applicable scenarios of static scheduling are greatly limited, and it is difficult to use static scheduling alone meet actual needs

Images